Remote chronic disease online follow-up information management method and system

Abstract

The present invention relates to the field of medical information management technology, and in particular to a method and system for managing chronic disease online follow-up information based on remote control. The method first collects and organizes the patient's historical medical record information, and encrypts and saves it using blockchain technology; secondly, by predicting the number of patients and historical medical record information within the outpatient doctor's schedule, the patient's appointment time period and doctor are dynamically recommended; when the patient confirms online, the diagnosis and treatment system notifies the doctor and sends the historical medical record information, and the doctor issues medical advice based on the patient's description and historical data, and generates a professional diagnosis and treatment report; finally, the method uses the generated professional diagnosis and treatment report to push health education materials and interactive activities to the patient to assist in rehabilitation treatment; the method not only improves the diagnosis and treatment efficiency and optimizes the patient experience through data-driven and intelligent recommendation, but also provides convenient and efficient online diagnosis and treatment services for chronic disease patients; it also realizes the optimal allocation of medical resources and personalized treatment of patients.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information management, and in particular to a method and system for remote online follow-up information management of chronic diseases. Background Art

[0002] With the rapid development of information technology and the continuous improvement of the medical system, telemedicine has become an important part of medical services. Especially for patients with chronic diseases, who require long-term disease monitoring and management, remote follow-up information management methods show great potential in this regard. However, the current remote chronic disease follow-up information management methods still have some shortcomings; first, data integration is not efficient enough, which makes it difficult for doctors to obtain complete medical history information of patients. Secondly, information security is insufficient, and patient privacy may be threatened. In addition, the existing system has limited intelligence and lacks automatic diagnosis suggestions and disease prediction functions. Finally, patient participation is not high and there is a lack of interactive mechanism, which affects the effective use of data. Therefore, a more advanced method is needed to solve these problems, improve the efficiency and security of follow-up information management, and increase patient participation and satisfaction.

[0003] For example, a Chinese patent with publication number CN115527666A discloses a medical record quality control and information management system and method for digital remote consultation, including a consultation communication network, a consultation medical terminal, an information management terminal, a client terminal, and an information comprehensive management server, wherein the consultation medical terminal, the information management terminal, the client terminal, and the information comprehensive management server all establish data connections with the consultation communication network, each consultation medical terminal and at least one information management terminal form a working group through the consultation communication network, and the distance between the consultation medical terminal and the information management terminal in the same working group is no more than 3 meters.

[0004] For example, a Chinese patent with publication number CN117238527A discloses a medical information management method and system based on big data. Through long short-term memory networks and attention mechanisms, features are extracted from medical consultation dialogues and saliency representation vectors are generated. In addition, the method pushes corresponding medical content information to users based on predicted user medical intention information, and updates network parameters through cross-domain cost parameters.

[0005] Through the analysis of the above-mentioned prior art, it is found that the prior art still has some shortcomings: 1) Data integration is not efficient enough, which makes it difficult for doctors to obtain the patient's complete medical history information; 2) Information security is insufficient and the patient's privacy may be threatened; 3) The intelligence level of the existing system is limited, and there is a lack of automatic diagnosis suggestions and condition assessment functions; 4) Patient participation is not high and there is a lack of interactive mechanism, which makes it impossible for chronic disease patients to timely understand the development of their own diseases; In order to solve the above problems, the present invention provides a remote-based chronic disease online follow-up information management method and system. Summary of the invention

[0006] The technical problem to be solved by the present invention is to address the deficiencies of the existing technology and provide a remote online follow-up information management method and system for chronic diseases. The method first collects and organizes the patient's historical medical record information, and encrypts and saves it using blockchain technology; secondly, by predicting the outpatient doctor's schedule and the number of patients, the patient's appointment time period and doctor are dynamically recommended; when the patient confirms online, the diagnosis and treatment system notifies the doctor and sends the historical medical record information. The doctor issues medical orders based on the patient's description and historical data and generates a professional diagnosis and treatment report; finally, the method uses the generated professional diagnosis and treatment report to push health education materials and interactive activities to the patient to assist in rehabilitation treatment; the method not only improves the diagnosis and treatment efficiency and optimizes the patient experience through data-driven and intelligent recommendations, but also provides convenient and efficient online diagnosis and treatment services for chronic disease patients; it also realizes the optimal allocation of medical resources and personalized treatment of patients.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The method for managing information of online follow-up consultation of chronic diseases based on remote treatment includes the following steps:

[0009] S1: Build and register to log in to the telemedicine online diagnosis and treatment system, and collect and organize the first outpatient examination data, past medical history, allergy history, previous outpatient records, medical image examinations and patient current status description data of chronic disease patients through the constructed telemedicine online diagnosis and treatment system, and use text matching, keyword extraction and image inspection strategies to extract key information, automatically match and verify, and perform medical image analysis on the patient's past medical history, allergy history, previous outpatient records, medical image examinations and current status description data input by the patient, and obtain the patient's historical medical record information data. At the same time, use blockchain technology to encrypt, save and regularly back up the obtained patient's historical medical record information data;

[0010] S2: Collect and organize the doctor schedules of various chronic disease clinics and the number of patients in each time period within the schedule through hospital data, and predict the number of patients in each time period within the schedule of all doctors on duty in various chronic disease clinics in the next 24 hours based on the collected schedules and the number of patients in the corresponding time periods;

[0011] S3: Based on the patient's historical medical record information data obtained in S1, the matching strategy is used to obtain the department clinic corresponding to the patient's symptoms, and a preliminary assessment of the patient's current symptom level is performed. At the same time, based on the predicted number of visits in each time period within the scheduling time of each doctor in the corresponding department clinic within 24 hours obtained in S2, the first outpatient examination data and the preliminary assessment results of the symptom level, the time period and doctor for appointment diagnosis are dynamically recommended to the patient;

[0012] S4: When the patient confirms the appointment time and doctor information online, the online diagnosis and treatment system automatically notifies the corresponding doctor and sends the generated patient historical medical record information data to the appointment doctor. When the appointment time is reached, the appointment doctor issues specific medical instructions based on the patient's remote description of symptoms and the patient's historical medical record information data sent by the system. The online diagnosis and treatment system generates a professional diagnosis and treatment report based on the specific medical instructions issued by the doctor and the patient's historical medical record information data, and pushes health education materials and health interaction activities for the corresponding disease to the patient based on the diagnosis and treatment report to assist the patient in rehabilitation treatment;

[0013] In S2, the numerical prediction of the number of patients in each time period of all doctors on duty in each chronic disease clinic in the next 24 hours adopts a segmented prediction strategy. The specific steps include:

[0014] S401, the data of the doctor schedules of chronic disease clinics collected and sorted by the hospital data and the number of patients in each time period in the schedule are divided into segments with 2 hours as a time period, and the divided data are deduplicated and de-noised to obtain the segmented pre-processed number of patients data of each doctor;

[0015] S402, constructing a segmented patient number prediction model with time hysteresis and seasonality based on the Prophet algorithm, and inputting the segmented pre-processed patient number data of each doctor into the constructed segmented patient number prediction model with hysteresis and seasonality, and calculating the patient number value of each on-duty doctor in each time period in the next 24-hour shift schedule;

[0016] In S3, a fuzzy evaluation strategy is used to conduct a preliminary assessment of the patient's current symptom level. The specific steps include:

[0017] S501. Constructing assessment factor sets for different types of chronic diseases based on previous outpatient records, medical image examinations, and patient current status description data, and setting assessment levels corresponding to the types of chronic diseases;

[0018] S502, constructing a comprehensive symptom level assessment model based on the grey correlation analysis method, and inputting the constructed assessment factor set of different types of chronic diseases and their corresponding assessment levels into the comprehensive symptom level assessment model, and calculating the assessment level of the symptom corresponding to the current chronic disease of each patient;

[0019] S503, obtaining a corresponding appointment recommendation priority score according to the obtained evaluation level of the current chronic disease corresponding to the symptom;

[0020] Dynamically recommend appointment time slots and doctors to patients, using a reinforcement learning strategy. The specific steps include:

[0021] S601, setting the input state s at the current time t according to the obtained segmented pre-processing data of the number of patients for each doctor, the evaluation level of the current chronic disease corresponding to the symptom and the appointment recommendation priority score t =(b i ,c j ,f j ,p i ,p ki调 ,p g ), where b i represents the number of appointments for the i-th time period, c j represents the jth disease assessment level of the current chronic disease of the scheduled patient, f j represents the appointment recommendation priority score corresponding to the jth disease assessment level, p i represents the probability that the patient cancels the appointment after successfully making the appointment in the i-th time period, p ki调 represents the probability that the kth doctor takes leave on the day of the i-th time period; p g Indicates the probability that the doctor booked is of the gth professional title; the action to be executed is set to a according to the input state t =(a1,a2), where a1 represents the execution action of successful appointment, and the online diagnosis and treatment system directly sends the appointment information to the patient; a2 represents the execution action of unsuccessful appointment, and the online diagnosis and treatment system adjusts the appointment time period, sends the adjusted appointment time period to the patient and confirms whether to adjust; set the combined reward function R according to the set input state t , specifically:

[0022]

[0023] Among them, r1 represents the reward function corresponding to the number of appointments, b 阈represents the maximum number of patients that can be seen by the k-th doctor during the appointment time period; r2 represents the reward function corresponding to the recommendation priority score; r3 represents the reward function corresponding to the patient canceling the appointment; r4 represents the reward function corresponding to the doctor taking leave on the appointment day; r5 represents the reward function corresponding to a successful appointment; r6 represents the reward function for the patient with the j-th disease to make an appointment with the g-th doctor; k1, k2, k3, k4, k5, k6 represent the gain coefficients of the corresponding reward functions, and α1, α2, α3, α4, α5, α6 represent the weighting coefficients of the corresponding reward functions;

[0024] S602: Build an intelligent appointment recommendation model based on the DDPG algorithm, and convert the input state s at the current time t built in S601 into t =(b i ,c j ,f j ,p i ,p ki调 ,p g ), the action performed is a t =(a1,a2) and the combined reward function R t Input the data into the constructed intelligent appointment recommendation model for training and optimization, and deploy the trained intelligent appointment recommendation model into the online diagnosis and treatment system;

[0025] S603, intelligently recommending appointment time periods and doctors to patients based on the acquired patient medical history information data, the number of patients in each time period, the evaluation level of corresponding symptoms, and the appointment recommendation priority score.

[0026] Specifically, the automatic matching verification in S1 adopts a knowledge graph and semantic matching fusion strategy; the specific steps include:

[0027] S101. Collect and organize various chronic disease names and their corresponding professional information through crawler technology, and use the chronic disease names and their corresponding professional information to construct corresponding knowledge graph triples, and import the constructed knowledge graph triples into the Neo4j database for storage;

[0028] S102, collect text data describing the patient's current status to construct input sentence A, and use various chronic disease names and their corresponding professional information in the knowledge graph to construct input sentence B, and input sentence A and input sentence B are simultaneously input into two parallel pre-trained Chinese Bert models for word embedding representation, and output semantic vector A and output semantic vector B at the same time;

[0029] S103, input the obtained output semantic vector A and output semantic vector B into two parallel BiLstm models at the same time to capture the contextual text features between the texts in the input semantic vector A and the semantic vector B respectively, obtain the corresponding output high-dimensional vector C and high-dimensional vector D, and input the obtained high-dimensional vector C and high-dimensional vector D into the corresponding pooling layer A and pooling layer B in parallel to obtain the text vector u and the text vector v;

[0030] S104, concatenate the obtained text vector u and text vector v to obtain fused text features, input the fused text features into the Softmax function, and output the text matching probability value; at the same time, use the text vector u and the text vector v to calculate the text matching distance, and use the text matching distance to obtain the cross entropy loss function to train the S102-S104 process. The specific text matching formula is: Wherein, d(u,v) represents the text matching distance between the current vector u and the text vector v, and cosh represents the hyperbolic function;

[0031] S105: Setting text replacement threshold d 阈 When the obtained text matching output probability value is less than the text replacement threshold, the text information in input sentence B is used to replace the text information in input sentence A, otherwise the input sentence A remains unchanged.

[0032] Specifically, key information extraction adopts a multi-source data hybrid extraction strategy, and the specific steps include:

[0033] S201, collecting and arranging the remote video diagnosis and treatment data and medical record image data in the past medical history and previous outpatient records collected by the online telemedicine diagnosis and treatment system; and extracting the video frames of the remote video diagnosis and treatment data through the openCV library to obtain the video frame image data;

[0034] S202: Using the OCR text extraction model to extract text sequences from the acquired video frame image data and medical record image data, and performing deduplication and denoising processing on the extracted patient history medical record text information sequence to obtain a preprocessed patient history medical record text information sequence;

[0035] S203. An improved multi-feature TextRank text keyword extraction model is constructed based on the TextRank model. At the same time, various chronic disease names and their corresponding professional medical record information stored in the Neo4j database are used to construct training sets and validation sets in a ratio of 8:2. The constructed training sets and validation sets are annotated using the BIO ternary method. The annotated training sets and validation sets are input into the improved multi-feature TextRank text keyword extraction model for pre-training to obtain a keyword extraction model.

[0036] S204. The obtained pre-processed patient historical medical record text information sequence is input into the keyword extraction model to extract the chronic disease name and the corresponding key doctor diagnosis information, and the obtained chronic disease name and the corresponding key doctor diagnosis information are constructed into triple structured data and saved in the Neo4j database.

[0037] Specifically, medical image analysis adopts an image text generation strategy, and the specific steps include:

[0038] S301, constructing image-text input pair data using the collected medical image examination data and the corresponding image doctor professional diagnosis text data in the outpatient records;

[0039] S302, constructing an image-text generation model, inputting the constructed image-text input pair data into the constructed image-text generation model, setting the cross entropy loss function threshold to 0.1, performing optimization training on the image-text generation model, and obtaining a trained image-text generation model;

[0040] S303. When a patient uploads medical image examination data to the telemedicine online diagnosis and treatment system, the system calls the trained image-text generation model to perform detection and analysis on the uploaded medical image examination data, and generates a preliminary medical image text detection report. At the same time, the system uses the automatic matching and verification strategy to match and verify the generated report, thereby generating a professional medical image text detection report, and then transmits the generated professional medical image text detection report and the corresponding pictures to the corresponding appointment doctor through the diagnosis and treatment system.

[0041] Specifically, the remote chronic disease online follow-up information management system includes a patient information management module, an intelligent follow-up appointment module, an online follow-up and intelligent auxiliary diagnosis module, a health education and interaction module, and a data security and privacy protection module;

[0042] Patient information management module, which is used to collect, store and update patients’ personal information, medical history and allergy history, and automatically parse the collected information and capture key information through natural language processing technology;

[0043] The intelligent follow-up appointment module is used to combine the patient's historical follow-up data and the doctor's work schedule to intelligently recommend the time and doctor for the follow-up visit, thus reducing the patient's waiting time;

[0044] Online follow-up and intelligent auxiliary diagnosis modules are used to realize online communication between patients and doctors. At the same time, intelligent algorithms and medical knowledge bases are used to provide auxiliary diagnosis suggestions to help doctors quickly and accurately understand the patient's condition and develop personalized diagnosis and treatment plans;

[0045] The health education and interaction module is used to provide patients with health education materials on chronic diseases and to enable online interactive Q&A between patients and doctors;

[0046] The data security and privacy protection module is used to combine blockchain technology to encrypt and back up patient personal information and medical record information, fully protecting the privacy and integrity of patient data.

[0047] Specifically, the patient information management module includes a patient registration and login unit, a medical history and allergy history recording unit, and a data processing unit; the intelligent follow-up appointment module includes a follow-up time recommendation unit and an appointment confirmation and notification unit; the online follow-up and intelligent auxiliary diagnosis module includes an online communication unit, a medical image analysis unit, and an intelligent auxiliary diagnosis unit;

[0048] The patient registration and login unit is used to realize the patient's online account registration and login and fill in personal basic information; the medical history and medical record collection unit is used to collect and store the patient's past medical history, allergy history, historical outpatient records and medical image examination data; the data processing unit is used to pre-process the collected patient's past medical history, allergy history, historical outpatient records and medical image examination data, capture and verify key information and perform medical image examination analysis; the follow-up time recommendation unit is used to intelligently recommend follow-up time and follow-up doctor based on the patient's historical follow-up data, the severity of the disease and the doctor's work schedule; the appointment confirmation and notification unit is used to automatically notify the appointment doctor via SMS or email after the patient confirms the appointment information online; the online communication unit is used for the patient to communicate with the doctor online to describe the current condition; the medical image analysis unit is used to assist in the analysis of the medical image examination data uploaded by the patient and provide an analysis and examination report; the intelligent auxiliary diagnosis unit is used to generate an intelligent auxiliary diagnosis recommendation report based on the patient's historical medical record data and the doctor's diagnosis results to assist the patient in subsequent treatment.

[0049] Specifically, the health education and interaction module includes a health education data push unit and an online interaction unit; the data security and privacy protection module includes a data encryption and protection unit and a blockchain data storage and verification unit; the health education data push unit is used to collect, organize, store and chronic disease health education data, and regularly push corresponding chronic disease health education data to patients to assist the patient's recovery process; the online interaction unit is used to establish a health question-and-answer interactive platform between patients and experts, and answer patients' questions; the data encryption and protection unit is used to encrypt the collected patient data and perform regular data backup; the blockchain data storage and verification unit is used to build a tamper-proof data storage and verification mechanism to improve the security and credibility of patient medical records.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. In view of the deficiencies in the prior art, the present invention first adopts a multi-source data mixed extraction strategy to extract different types of patient medical record information from multiple sources. Secondly, by combining the knowledge graph and semantic matching fusion strategy, the extracted patient input information is matched and corrected, thereby improving the integration and professionalism of the patient's medical record data; thirdly, the present invention also uses the hospital database information to calculate the number of doctors' visits in each time period, evaluate the level of the patient's current symptoms, and determine the priority of appointment recommendations accordingly; this method further enhances the system's ability to integrate doctor and patient medical record information and improves the accuracy of appointment recommendations; fourthly, the present invention also uses a reinforcement learning algorithm to make intelligent decisions based on the acquired patient medical record information, number of patients, doctor's professional title, current symptom level and recommendation priority. This not only improves the success rate of appointments, but also makes intelligent appointments more targeted, ensuring that patients with serious conditions can make appointments with doctors with higher professional titles more quickly.

[0052] 2. The present invention aims to solve the problems of low patient participation and lack of interactive mechanism by generating professional diagnosis and treatment reports and regularly pushing health education materials and interactive activities related to their diseases to patients, thereby effectively improving the participation of patients in the whole treatment process; this process not only enhances the patient's understanding of their own diseases, but also promotes communication between patients and medical teams through interactive activities, thereby enhancing the superiority of the information management system in assisting patients in rehabilitation treatment;

[0053] 3. In order to solve the problem of insufficient information security, the present invention uses blockchain technology to encrypt and save the acquired patient historical medical record information data and perform regular backup, which not only improves the security of patient data, but also enhances the trust and reliability of medical information systems in data processing and storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0055] Figure 1 This is a flow chart of a method for managing chronic disease online follow-up information based on remote control in Example 1 of the present invention;

[0056] Figure 2 This is a knowledge graph and semantic matching fusion strategy structure diagram of Example 1 of the present invention;

[0057] Figure 3 This is a structural diagram of a multi-source data hybrid extraction strategy in Example 1 of the present invention;

[0058] Figure 4 This is a module architecture diagram of a remote online follow-up information management system for chronic diseases according to Example 2 of the present invention. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0060] Embodiment 1:

[0061] For this purpose, the present invention provides an embodiment: a method for managing chronic disease online follow-up information based on remote Figure 1 , the specific steps include:

[0062] S1: Build and register to log in to the telemedicine online diagnosis and treatment system, and collect and organize the first outpatient examination data, past medical history, allergy history, previous outpatient records, medical image examinations and patient current status description data of chronic disease patients through the constructed telemedicine online diagnosis and treatment system, and use text matching, keyword extraction and image inspection strategies to extract key information, automatically match and verify, and perform medical image analysis on the patient's past medical history, allergy history, previous outpatient records, medical image examinations and current status description data input by the patient, and obtain the patient's historical medical record information data. At the same time, use blockchain technology to encrypt and save the obtained patient's historical medical record information data and regularly back up it; specifically, previous outpatient records include the first outpatient examination data and subsequent on-site examination records in the hospital and remote follow-up diagnosis records;

[0063] S2: Collect and organize the doctor schedules of various chronic disease clinics and the number of patients in each time period within the schedule through hospital data, and predict the number of patients in each time period within the schedule of all on-duty doctors in various chronic disease clinics in the next 24 hours based on the collected schedules and the number of patients in the corresponding time periods;

[0064] S3: Based on the patient's historical medical record information data obtained in S1, the matching strategy is used to obtain the department clinic corresponding to the patient's symptoms, and a preliminary assessment of the patient's current symptom level is performed. At the same time, based on the predicted number of patients in each time period within the scheduling time of each doctor in the corresponding department clinic within 24 hours, the first outpatient examination data and the preliminary assessment results of the symptom level predicted in S2, the time period and doctor for appointment diagnosis are dynamically recommended to the patient;

[0065] S4: When the patient confirms the appointment time period and doctor information online, the online diagnosis and treatment system automatically notifies the corresponding doctor and sends the generated patient historical medical information data to the appointment doctor. When the appointment time period is reached, the appointment doctor issues specific medical instructions based on the patient's remote description of symptoms and the patient's historical medical information data sent by the system. The online diagnosis and treatment system generates a professional diagnosis and treatment report based on the specific medical instructions issued by the doctor and the patient's historical medical information data, and pushes health education materials and health interaction activities for the corresponding disease to the patient based on the diagnosis and treatment report to assist the patient in rehabilitation treatment.

[0066] This method improves the convenience and participation of patients in diagnosis and treatment by collecting historical medical information of patients, predicting doctor schedules and the number of patients, and dynamically recommending appointment times. It combines blockchain technology to ensure information security, push health education and interactive activities, and enhance the rehabilitation effect of patients. The entire process effectively integrates medical resources and improves the quality of medical services and patient satisfaction.

[0067] Specifically, see Figure 2 , automatic matching verification in S1 adopts a knowledge graph and semantic matching fusion strategy; the specific steps include:

[0068] S101. Collect and organize various chronic disease names and their corresponding professional information through crawler technology, and use the chronic disease names and their corresponding professional information to construct corresponding knowledge graph triples, and import the constructed knowledge graph triples into the Neo4j database for storage; specifically, the specific form of the construction of the triples is diabetes-symptoms-polyuria, diabetes-treatment methods-drug therapy, diet control, exercise therapy, etc. Other diseases and symptoms and other medical record information are constructed according to the above method;

[0069] S102, collect the text data describing the patient's current status to construct input sentence A, and use the names of various chronic diseases in the knowledge graph and their corresponding professional information to construct input sentence B, and input sentence A and input sentence B are simultaneously input into two parallel pre-trained Chinese Bert models for word embedding representation, and output semantic vector A and output semantic vector B at the same time; specifically, set the Bert batch input size to 512 and the length of the input sequence to 64;

[0070] S103, input the obtained output semantic vector A and output semantic vector B into two parallel BiLstm models at the same time to capture the contextual text features between the texts in the input semantic vector A and the semantic vector B respectively, obtain the corresponding output high-dimensional vector C and high-dimensional vector D, and input the obtained high-dimensional vector C and high-dimensional vector D into the corresponding pooling layer A and pooling layer B in parallel to obtain the text vector u and the text vector v;

[0071] S104, concatenate the obtained text vector u and text vector v to obtain fused text features, input the fused text features into the Softmax function, and output the text matching probability value; at the same time, use the text vector u and the text vector v to calculate the text matching distance, and use the text matching distance to obtain the cross entropy loss function to train the S102-S104 process. The specific text matching formula is: Wherein, d(u,v) represents the text matching distance between the current vector u and the text vector v, and cosh represents the hyperbolic function;

[0072] S105: Setting text replacement threshold d 阈 When the obtained text matching output probability value is less than the text replacement threshold, the text information in the input sentence B is used to replace the text information in the input sentence A, otherwise the input sentence A remains unchanged. Specifically, the text replacement threshold d is set. 阈 is 0.95.

[0073] This method improves the accuracy of automatic matching verification of patient status descriptions by integrating knowledge graphs and semantic matching strategies; it captures the contextual features between texts and enhances the representation ability of semantic vectors by combining the BERT model and the BiLSTM model; and it achieves precise matching and replacement by calculating text matching probability and distance, thereby improving the accuracy of patient information extraction and providing strong support for subsequent diagnosis and treatment decisions.

[0074] Specifically, see Figure 3 ,Key information extraction adopts a multi-source data hybrid extraction strategy, and the specific steps include:

[0075] S201, collecting and arranging the remote video diagnosis and treatment data and medical record image data in the past medical history and previous outpatient records collected by the online telemedicine diagnosis and treatment system; and extracting the video frames of the remote video diagnosis and treatment data through the openCV library to obtain the video frame image data;

[0076] S202: Using the acquired video frame image data and medical record image data respectively using an OCR text extraction model to extract text sequences, and performing deduplication and denoising processing on the extracted patient history medical record text information sequence to obtain a pre-processed patient history medical record text information sequence; specifically, the OCR text extraction model used in this embodiment uses the open source TesseractOCR pre-trained model developed by Google;

[0077] S203. An improved multi-feature TextRank text keyword extraction model is constructed based on the TextRank model. At the same time, various chronic disease names and their corresponding professional medical record information saved in the Neo4j database are used to construct training sets and validation sets in a ratio of 8:2. The constructed training sets and validation sets are annotated using the BIO ternary method, and the annotated training sets and validation sets are input into the improved multi-feature TextRank text keyword extraction model for pre-training to obtain a keyword extraction model. Specifically, the multi-feature TextRank text keyword extraction model is obtained by inputting various chronic disease names and their corresponding professional medical record information to construct medical record keyword graph features, chronic disease names and corresponding symptom keyword statistical features, and chronic disease names and corresponding symptom keyword position features. The three features are input into three TextRank models in parallel for text feature extraction, and the three extracted text features are cascaded to obtain fused text features, and the fused text features are input into two fully connected layers to output the probability of keyword extraction.

[0078] S204. The obtained pre-processed patient historical medical record text information sequence is input into the keyword extraction model to extract the chronic disease name and the corresponding key doctor diagnosis information, and the obtained chronic disease name and the corresponding key doctor diagnosis information are constructed into triple structured data and saved in the Neo4j database.

[0079] The above strategy can effectively extract the names of chronic diseases and key diagnostic information, and save them in a structured manner in the Neo4j database, which improves the accuracy and efficiency of information extraction and provides strong support for medical decision-making.

[0080] Specifically, medical image analysis adopts an image text generation strategy, and the specific steps include:

[0081] S301, constructing image-text input pair data using the collected medical image examination data and the corresponding image doctor professional diagnosis text data in the outpatient records;

[0082] S302, construct an image-text generation model, input the constructed image-text input pair data into the constructed image-text generation model, set the cross entropy loss function threshold to 0.1, optimize and train the image-text generation model, and obtain the trained image-text generation model; specifically, the image-text generation model uses a pre-trained resnet101 algorithm to extract features from the input medical image examination data to obtain medical image features, uses a pre-trained Chinese algorithm to extract features from the corresponding image doctor's professional diagnosis text data to obtain image-text features, and uses a collaborative attention network to fuse the obtained medical image features and image-text features, calculates the fused features, and inputs the fused features into the fully connected layer to obtain the final image-text output accuracy;

[0083] S303. When a patient uploads medical image examination data to the telemedicine online diagnosis and treatment system, the system calls the trained image-text generation model to perform detection and analysis on the uploaded medical image examination data, and generates a preliminary medical image text detection report. At the same time, the system uses the automatic matching and verification strategy to match and verify the generated report, thereby generating a professional medical image text detection report, and then transmits the generated professional medical image text detection report and the corresponding pictures to the corresponding appointment doctor through the diagnosis and treatment system.

[0084] The above medical image analysis process integrates medical images and doctors' professional diagnosis text data to construct image-text input pair data, extracts image and text features through pre-training algorithms, and then uses collaborative attention networks to fuse these features, thereby improving the accuracy of the model's understanding of medical image content. In telemedicine online diagnosis and treatment systems, this strategy can quickly generate preliminary medical image text detection reports, and improve the professionalism of the reports through automatic matching and verification strategies, providing doctors with valuable reference information, optimizing the diagnosis and treatment process, and improving patient experience.

[0085] Specifically, the number of patients visiting each chronic disease clinic in each time period within the scheduled time of all doctors on duty in S2 in the next 24 hours adopts a segmented prediction strategy. The specific steps include:

[0086] S401, the data of the doctor schedules of chronic disease clinics collected and sorted by the hospital data and the number of patients in each time period in the schedule are divided into segments with 2 hours as a time period, and the divided data are deduplicated and de-noised to obtain the segmented pre-processed number of patients data of each doctor;

[0087] S402. Construct a segmented patient number prediction model with time hysteresis and seasonality based on the Prophet algorithm, and input the segmented pre-processed patient number data of each doctor into the constructed segmented patient number prediction model with hysteresis and seasonality to calculate the number of patients in each time period of each on-duty doctor's schedule for the next 24 hours.

[0088] Specifically, in S3, the patient's current symptom level is preliminarily evaluated using a fuzzy evaluation strategy. The specific steps include:

[0089] S501. Construct a set of assessment factors for different types of chronic diseases based on the outpatient records, medical image examinations, and the patient's current status description data, and set the assessment level corresponding to the chronic disease type; specifically, the assessment level corresponding to the chronic disease type is obtained through a professional medical information database, such as pneumonia including mild pneumonia, moderate pneumonia, and severe pneumonia;

[0090] S502, constructing a comprehensive symptom level assessment model based on the grey correlation analysis method, and inputting the constructed assessment factor set of different types of chronic diseases and their corresponding assessment levels into the comprehensive symptom level assessment model, and calculating the assessment level of the symptom corresponding to the current chronic disease of each patient;

[0091] S503. Obtain the corresponding appointment recommendation priority score according to the evaluation level of the symptoms corresponding to the current chronic disease. The appointment recommendation priority score is obtained by assigning points based on the evaluation level of the symptoms corresponding to the current chronic disease, such as 1 point for mild pneumonia, 2 points for moderate pneumonia, and 3 points for severe pneumonia. Other types of chronic diseases are assigned points in ascending order according to the severity of the symptoms as described above.

[0092] This fuzzy evaluation strategy constructs a comprehensive evaluation factor set by comprehensively considering the patient's outpatient records, medical image examinations and current status, and builds an evaluation model based on the grey correlation analysis method to ensure the accuracy and objectivity of the evaluation results. At the same time, the evaluation level is converted into an appointment recommendation priority score to provide decision support for medical institutions, optimize resource allocation, and improve the efficiency and quality of medical services.

[0093] Specifically, a reinforcement learning strategy is used to dynamically recommend appointment time periods and doctors to patients. The specific steps include:

[0094] S601, setting the input state s at the current time t according to the obtained segmented pre-processing data of the number of patients for each doctor, the evaluation level of the current chronic disease corresponding to the symptom and the appointment recommendation priority score t =(b i ,c j ,f j ,p i ,p ki调 ,pg ), where b i represents the number of appointments for the i-th time period, c j represents the jth disease assessment level of the current chronic disease of the scheduled patient, f j represents the appointment recommendation priority score corresponding to the jth disease assessment level, p i represents the probability that the patient cancels the appointment after successfully making an appointment in the i-th time period, which is calculated based on the historical appointment records; p ki调 represents the probability that the kth doctor takes leave on the day of the i-th time period, which is calculated based on the doctor's historical leave records and work adjustment records; p g Indicates the probability that the doctor booked is of the gth professional title. Specifically, the professional titles of doctors include resident physician, attending physician, associate chief physician, and chief physician. This probability is calculated based on the corresponding professional title records of the patient's historical appointments with the doctor. Set the action to be executed according to the input status as a t =(a1,a2), where a1 represents the execution action of successful appointment, and the online diagnosis and treatment system directly sends the appointment information to the patient; a2 represents the execution action of unsuccessful appointment, and the online diagnosis and treatment system adjusts the appointment time period, sends the adjusted appointment time period to the patient and confirms whether to adjust; set the combined reward function R according to the set input state t , specifically:

[0095]

[0096] Among them, r1 represents the reward function corresponding to the number of appointments, b 阈 represents the maximum number of patients that can be seen by the k-th doctor during the appointment time period; r2 represents the reward function corresponding to the recommendation priority score; r3 represents the reward function corresponding to the patient canceling the appointment; r4 represents the reward function corresponding to the doctor taking leave on the appointment day; r5 represents the reward function corresponding to a successful appointment; r6 represents the reward function for the patient with the j-th disease to make an appointment with the g-th doctor; k1, k2, k3, k4, k5, k6 represent the gain coefficients of the corresponding reward functions, and α1, α2, α3, α4, α5, α6 represent the weighting coefficients of the corresponding reward functions;

[0097] S602: Build an intelligent appointment recommendation model based on the DDPG algorithm, and convert the input state s at the current time t built in S601 into t =(b i ,c j ,f j ,p i ,p ki调 ,p g ), the action performed is a t=(a1,a2) and the combined reward function R t Input the data into the constructed intelligent appointment recommendation model for training and optimization, and deploy the trained intelligent appointment recommendation model into the online diagnosis and treatment system;

[0098] S603. Intelligently recommend appointment time periods and doctors to patients based on the acquired patient medical history information data, the number of patients in each time period, the evaluation level of the corresponding symptoms, and the appointment recommendation priority score. Specifically, through the above recommendation process, the system will recommend the most suitable doctor to the patient based on the acquired patient medical history information, the number of patients, and the level of symptoms. For example, if patient A suffers from severe pneumonia, the system will recommend a doctor with a fewer number of patients and a senior professional title and a time period to the patient, so that the patient can receive diagnosis and treatment faster and better.

[0099] This process ensures that recommendations are more accurate and personalized by comprehensively considering the doctor's working status, the patient's disease assessment level and appointment priority, thereby improving the patient's medical experience. Secondly, the reinforcement learning strategy enables the system to continuously learn and optimize, continuously improve the efficiency and accuracy of appointment recommendations, and achieve a reasonable allocation of medical resources. Finally, this strategy helps to improve the level of intelligence of medical services, enhance the flexibility and responsiveness of the medical system, and provide patients with more convenient and efficient diagnosis and treatment services.

[0100] Embodiment 2:

[0101] See also Figure 4 , the present invention provides an embodiment: a remote-based online follow-up information management system for chronic diseases, including a remote-based online follow-up information management system for chronic diseases, including a patient information management module, an intelligent follow-up appointment module, an online follow-up and intelligent auxiliary diagnosis module, a health education and interaction module, and a data security and privacy protection module;

[0102] The patient information management module is used to collect, store and update the patient's personal information, medical history and allergy history information, and automatically parse the collected information and capture key information through natural language processing technology; the intelligent follow-up appointment module is used to combine the patient's historical follow-up data and the doctor's work schedule to intelligently recommend the time and doctor for follow-up, thereby reducing the patient's waiting time; the online follow-up and intelligent auxiliary diagnosis module is used to realize online communication between patients and doctors, and at the same time use intelligent algorithms and medical knowledge bases to provide auxiliary diagnosis suggestions to help doctors quickly and accurately understand the patient's condition and formulate personalized diagnosis and treatment plans; the health education and interaction module is used to provide patients with health education materials on chronic diseases, and at the same time realize online interactive questions and answers between patients and doctors; the data security and privacy protection module is used to combine blockchain technology to encrypt and back up patients' personal information and medical records information data, and comprehensively protect the privacy and integrity of patient data.

[0103] The patient information management module includes a patient registration and login unit, a medical history and allergy history recording unit, and a data processing unit; the intelligent follow-up appointment module includes a follow-up time recommendation unit and an appointment confirmation and notification unit; the online follow-up and intelligent auxiliary diagnosis module includes an online communication unit, a medical image analysis unit, and an intelligent auxiliary diagnosis unit;

[0104] The patient registration and login unit is used to realize the patient's online account registration and login and fill in personal basic information; the medical history and medical record collection unit is used to collect and store the patient's past medical history, allergy history, historical outpatient records and medical image examination data; the data processing unit is used to pre-process the collected patient's past medical history, allergy history, historical outpatient records and medical image examination data, capture and verify key information and perform medical image examination analysis; the follow-up time recommendation unit is used to intelligently recommend follow-up time and follow-up doctor based on the patient's historical follow-up data, the severity of the disease and the doctor's work schedule; the appointment confirmation and notification unit is used to automatically notify the appointment doctor via SMS or email after the patient confirms the appointment information online; the online communication unit is used for the patient to communicate with the doctor online to describe the current condition; the medical image analysis unit is used to assist in the analysis of the medical image examination data uploaded by the patient and provide an analysis and examination report; the intelligent auxiliary diagnosis unit is used to generate an intelligent auxiliary diagnosis recommendation report based on the patient's historical medical record data and the doctor's diagnosis results to assist the patient in subsequent treatment.

[0105] The health education and interaction module includes a health education data push unit and an online interaction unit; the data security and privacy protection module includes a data encryption and protection unit and a blockchain data storage and verification unit; the health education data push unit is used to collect, organize, store and chronic disease health education data, and regularly push corresponding chronic disease health education data to patients to assist in the patient's recovery process; the online interaction unit is used to establish a health question-and-answer interactive platform between patients and experts and answer patients' questions; the data encryption and protection unit is used to encrypt the collected patient data and perform regular data backup; the blockchain data storage and verification unit is used to build a tamper-proof data storage and verification mechanism to improve the security and credibility of patient medical records.

[0106] The specific workflow of the system is that the patient first registers and logs in online through the patient information management module, fills in personal information, past medical history, allergy history and medical image examination data; these information are pre-processed and key information is captured by the data processing unit, and then stored in the Neo4j database. Secondly, the intelligent follow-up appointment module intelligently recommends the follow-up time and doctor based on the patient's historical follow-up data, the severity of the disease and the doctor's work schedule; after the patient confirms the appointment information, the system will notify the doctor of the appointment by SMS or email; during the follow-up, the patient communicates with the doctor online through the online follow-up and intelligent auxiliary diagnosis module, describes the current condition, and uploads medical image examination data. The medical image analysis unit will assist the doctor in image analysis and provide an analysis report; at the same time, the intelligent auxiliary diagnosis unit combines the patient's historical medical record data and the doctor's diagnosis results to generate an intelligent auxiliary diagnosis recommendation report; in addition, the health education and interaction module will regularly push chronic disease health education data to patients, and provide an online interactive platform to answer patients' questions; during the entire operation of the follow-up information management system, the data security and privacy protection module ensures the security and privacy of patient data, and builds an unalterable data storage and verification mechanism through data encryption and blockchain technology.

[0107] The system has the following significant beneficial effects. First, it greatly improves the communication efficiency between patients and doctors. Through the online communication unit, patients can describe their condition in a timely manner and doctors can give feedback quickly. Secondly, the system provides doctors with powerful auxiliary diagnosis suggestions through the intelligent auxiliary diagnosis unit, which improves the accuracy and efficiency of diagnosis and treatment. In addition, the health education and interaction module helps patients better understand disease knowledge and promotes the recovery process. Finally, the data security and privacy protection module ensures the security and privacy of patient data and enhances patients' trust in the system. In summary, the system has brought many conveniences to patients and doctors and has a positive role in promoting chronic disease management.

[0108] Embodiment 3:

[0109] A computer-readable storage medium having computer instructions stored thereon, which, when executed, execute any of the above-mentioned methods for managing online remote chronic disease follow-up information.

[0110] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A remote chronic disease online follow-up information management method, characterized in that: The following steps are involved: S1: Build and register to log in to the telemedicine online diagnosis and treatment system, and collect and organize the first outpatient examination data, past medical history, allergy history, previous outpatient records, medical image examinations and patient current status description data of chronic disease patients through the constructed telemedicine online diagnosis and treatment system, and use text matching, keyword extraction and image inspection strategies to extract key information, automatically match and verify, and perform medical image analysis on the patient's past medical history, allergy history, previous outpatient records, medical image examinations and current status description data input by the patient, and obtain the patient's historical medical record information data. At the same time, use blockchain technology to encrypt, save and regularly back up the obtained patient's historical medical record information data; S2: Collect and organize the doctor schedules, corresponding professional titles and the number of patients in each time period of each chronic disease clinic through hospital data, and predict the number of patients in each time period of each chronic disease clinic in the next 24 hours according to the collected schedules and the number of patients in the corresponding time periods; S3: Based on the patient's historical medical record information data obtained in S1, the matching strategy is used to obtain the department clinic corresponding to the patient's symptoms, and a preliminary assessment of the patient's current symptom level is performed. At the same time, based on the predicted number of patients in each time period within the scheduling time of each doctor in the corresponding department clinic within 24 hours, the first outpatient examination data and the preliminary assessment results of the symptom level predicted in S2, the time period and doctor for appointment diagnosis are dynamically recommended to the patient; S4: When the patient confirms the appointment time and doctor information online, the online diagnosis and treatment system automatically notifies the corresponding doctor and sends the generated patient historical medical record information data to the appointment doctor. When the appointment time is reached, the appointment doctor issues specific medical advice based on the patient's remote description of symptoms and the patient's historical medical record information data sent by the system. The online diagnosis and treatment system generates a professional diagnosis and treatment report based on the specific medical advice issued by the doctor and the patient's historical medical record information data, and pushes health education materials and health interaction activities for the corresponding disease to the patient based on the diagnosis and treatment report to assist the patient in rehabilitation treatment; The prediction of the number of patients in each time period of all doctors on duty in each chronic disease clinic in the next 24 hours in S2 adopts a segmented prediction strategy, and the specific steps include: S401, the data of the doctor schedules of chronic disease clinics collected and sorted by the hospital data and the number of patients in each time period in the schedule are divided into segments with 2 hours as a time period, and the divided data are deduplicated and de-noised to obtain the segmented pre-processed number of patients data of each doctor; S402, constructing a segmented patient number prediction model with time hysteresis and seasonality based on the Prophet algorithm, and inputting the segmented pre-processed patient number data of each doctor into the constructed segmented patient number prediction model with hysteresis and seasonality, and calculating the number of scheduled patients in each time period of the next 24-hour shift schedule of each on-duty doctor; In S3, a fuzzy evaluation strategy is used to perform a preliminary assessment of the patient's current symptom level. The specific steps include: S501. Constructing assessment factor sets for different types of chronic diseases based on previous outpatient records, medical image examinations, and patient current status description data, and setting assessment levels corresponding to the types of chronic diseases; S502, constructing a comprehensive symptom level assessment model based on the grey correlation analysis method, and inputting the constructed assessment factor set of different types of chronic diseases and their corresponding assessment levels into the comprehensive symptom level assessment model, and calculating the assessment level of the symptom corresponding to the current chronic disease of each patient; S503, obtaining a corresponding appointment recommendation priority score according to the obtained evaluation level of the current chronic disease corresponding to the symptom; The method of dynamically recommending the time period and doctor for the appointment diagnosis to the patient adopts a reinforcement learning strategy, and the specific steps include: S601, setting the input state s at the current time t according to the obtained segmented pre-processing data of the number of patients for each doctor, the evaluation level of the current chronic disease corresponding to the symptom and the appointment recommendation priority score t =(b i ,c j ,f j ,p i ,p ki调 ,p g ), where b i represents the number of appointments for the i-th time period, c j represents the jth disease assessment level of the current chronic disease of the scheduled patient, f j represents the appointment recommendation priority score corresponding to the jth disease assessment level, p i represents the probability that the patient cancels the appointment after successfully making the appointment in the i-th time period, p ki调 represents the probability that the kth doctor takes leave on the day of the i-th time period; p g Indicates the probability that the doctor booked is of the gth professional title; the action to be executed is set to a according to the input state t =(a1,a2), where a1 represents the execution action of successful appointment, and the online diagnosis and treatment system directly sends the appointment information to the patient; a2 represents the execution action of unsuccessful appointment, and the online diagnosis and treatment system adjusts the appointment time period, sends the adjusted appointment time period to the patient and confirms whether to adjust; set the combined reward function R according to the set input state t , specifically: Among them, r1 represents the reward function corresponding to the number of appointments, b 阈 represents the maximum number of patients that the k-th doctor can see during the appointment time period; r2 represents the reward function corresponding to the recommendation priority score; r3 represents the reward function corresponding to the patient canceling the appointment; r4 represents the reward function corresponding to the doctor taking leave on the appointment day; r5 represents the reward function corresponding to a successful appointment; r6 represents the reward function for the patient with the j-th disease to make an appointment with the g-th doctor; k1, k2, k3, k4, k5, k6 represent the gain coefficients of the corresponding reward functions respectively, and α1, α2, α3, α4, α5, α6 represent the weighting coefficients of the corresponding reward functions respectively; S602: Build an intelligent appointment recommendation model based on the DDPG algorithm, and convert the input state s at the current time t built in S601 into t =(b i ,c j ,f j ,p i ,p ki调 ,p g ), the action performed is a t =(a1,a2) and the combined reward function R t Input the data into the constructed intelligent appointment recommendation model for training and optimization, and deploy the trained intelligent appointment recommendation model into the online diagnosis and treatment system; S603, intelligently recommending appointment time periods and doctors to patients based on the acquired patient medical history information data, the number of patients in each time period, the evaluation level of corresponding symptoms, and the appointment recommendation priority score.

2. The method for managing chronic disease online follow-up information based on remote control according to claim 1, characterized in that: The automatic matching verification in S1 adopts a knowledge graph and semantic matching fusion strategy; The specific steps include: S101. Collect and organize various chronic disease names and their corresponding professional information through crawler technology, and use the chronic disease names and their corresponding professional information to construct corresponding knowledge graph triples, and import the constructed knowledge graph triples into the Neo4j database for storage; S102, collect text data describing the patient's current status to construct input sentence A, and use various chronic disease names and their corresponding professional information in the knowledge graph to construct input sentence B, and input sentence A and input sentence B are simultaneously input into two parallel pre-trained Chinese Bert models for word embedding representation, and output semantic vector A and output semantic vector B at the same time; S103, input the obtained output semantic vector A and output semantic vector B into two parallel BiLstm models at the same time to capture the contextual text features between the texts in the input semantic vector A and the semantic vector B respectively, obtain the corresponding output high-dimensional vector C and high-dimensional vector D, and input the obtained high-dimensional vector C and high-dimensional vector D into the corresponding pooling layer A and pooling layer B in parallel to obtain the text vector u and the text vector v; S104, concatenate the obtained text vector u and text vector v to obtain fused text features, input the fused text features into the Softmax function, and output the text matching probability value; at the same time, use the text vector u and the text vector v to calculate the text matching distance, and use the text matching distance to obtain the cross entropy loss function to train the S102-S104 process. The specific text matching formula is: Wherein, d(u,v) represents the text matching distance between the current vector u and the text vector v, and cosh represents the hyperbolic function; S105: Setting text replacement threshold d 阈 When the obtained text matching output probability value is less than the text replacement threshold, the text information in input sentence B is used to replace the text information in input sentence A, otherwise the input sentence A remains unchanged.

3. The method for managing chronic disease online follow-up information based on remote control according to claim 2, characterized in that: The key information extraction adopts a multi-source data mixed extraction strategy, and the specific steps include: S201, collecting and arranging the remote video diagnosis and treatment data and medical record image data in the past medical history and previous outpatient records collected by the remote medical online diagnosis and treatment system; and extracting the video frames of the remote video diagnosis and treatment data through the openCV library to obtain video frame image data; S202: Using the OCR text extraction model to extract text sequences from the acquired video frame image data and medical record image data, and performing deduplication and denoising processing on the extracted patient history medical record text information sequence to obtain a preprocessed patient history medical record text information sequence; S203, constructing an improved multi-feature TextRank text keyword extraction model based on the TextRank model, and using the various chronic disease names and their corresponding professional medical record information stored in the Neo4j database to construct a training set and a validation set in a ratio of 8:2, and using the BIO ternary method to annotate the constructed training set and validation set, and input the annotated training set and validation set into the improved multi-feature TextRank text keyword extraction model for pre-training to obtain a keyword extraction model; S204. Input the obtained pre-processed patient historical medical record text information sequence into the keyword extraction model, extract the chronic disease name and the corresponding key doctor diagnosis information, and construct the obtained chronic disease name and the corresponding doctor key diagnosis information into triple structured data, and save it in the Neo4j database.

4. The method for managing chronic disease online follow-up information based on remote control according to claim 3 is characterized in that: The medical image analysis adopts an image text generation strategy, and the specific steps include: S301, constructing image-text input pair data using the collected medical image examination data and the doctor's professional diagnosis text data corresponding to the images in the previous outpatient records; S302, constructing an image-text generation model, inputting the constructed image-text input pair data into the constructed image-text generation model, setting the cross entropy loss function threshold to 0.1, performing optimization training on the image-text generation model, and obtaining a trained image-text generation model; S303. When a patient uploads medical image examination data to the telemedicine online diagnosis and treatment system, the system calls the trained image-text generation model to perform detection and analysis on the uploaded medical image examination data, and generates a preliminary medical image text detection report. At the same time, the system uses the automatic matching and verification strategy to match and verify the generated report, thereby generating a professional medical image text detection report, and then transmits the generated professional medical image text detection report and the corresponding pictures to the corresponding appointment doctor through the diagnosis and treatment system.

5. A remote chronic disease online follow-up information management system, which is implemented based on the remote chronic disease online follow-up information management method according to any one of claims 1 to 4, characterized in that: The system includes a patient information management module, an intelligent follow-up appointment module, an online follow-up and intelligent auxiliary diagnosis module, a health education and interaction module, and a data security and privacy protection module; The patient information management module is used to collect, store, and update the patient's personal information, medical history, and allergy history information, and automatically parse the collected information and capture key information through natural language processing technology; The intelligent follow-up appointment module is used to combine the patient's historical follow-up data and the doctor's work schedule to intelligently recommend the time and doctor for the follow-up visit, thereby reducing the patient's waiting time; The online follow-up and intelligent auxiliary diagnosis module is used to realize online communication between patients and doctors, and at the same time use intelligent algorithms and medical knowledge bases to provide auxiliary diagnosis suggestions, helping doctors to quickly and accurately understand the patient's condition and formulate personalized diagnosis and treatment plans; The health education and interaction module is used to provide patients with health education materials on chronic diseases and to enable online interactive Q&A between patients and doctors; The data security and privacy protection module is used to combine blockchain technology to encrypt and back up patient personal information and medical record information, thereby fully protecting the privacy and integrity of patient data.

6. The remote online follow-up information management system for chronic diseases according to claim 5 is characterized in that: The patient information management module includes a patient registration and login unit, a medical history and allergy history recording unit, and a data processing unit; the intelligent follow-up appointment module includes a follow-up time recommendation unit and an appointment confirmation and notification unit; the online follow-up and intelligent auxiliary diagnosis module includes an online communication unit, a medical image analysis unit, and an intelligent auxiliary diagnosis unit; The patient registration and login unit is used to realize the patient's online account registration and login and the filling of personal basic information; the medical history and medical record collection unit is used to collect and store the patient's past medical history, allergy history, historical outpatient records and medical image examination data; the data processing unit is used to pre-process the collected patient's past medical history, allergy history, historical outpatient records and medical image examination data, capture and verify key information and perform medical image examination analysis; the follow-up time recommendation unit is used to intelligently recommend follow-up time and follow-up doctor based on the patient's historical follow-up data, the severity of the disease and the doctor's work schedule; the appointment confirmation and notification unit is used to automatically notify the doctor of the appointment via SMS or email after the patient confirms the appointment information online; the online communication unit is used for the patient to communicate with the doctor online to describe the current condition; the medical image analysis unit is used to assist in the analysis of the medical image examination data uploaded by the patient and provide an analysis and examination report; The intelligent auxiliary diagnosis unit is used to generate an intelligent auxiliary diagnosis recommendation report based on the patient's historical medical record data and the doctor's diagnosis results to assist the patient in subsequent treatment.

7. The remote online follow-up information management system for chronic diseases according to claim 6 is characterized in that: The health education and interaction module includes a health education data push unit and an online interaction unit; the data security and privacy protection module includes a data encryption and protection unit and a blockchain data storage and verification unit; The health education data push unit is used to collect, organize and store chronic disease health education data, and regularly push the corresponding chronic disease health education data to the patient to assist the patient's rehabilitation process; The online interactive unit is used to establish a health question-and-answer interactive platform between patients and experts to answer patients' questions; The data encryption and protection unit is used to encrypt the collected patient data and perform regular data backup; The blockchain data storage and verification unit is used to build a tamper-proof data storage and verification mechanism to improve the security and credibility of patient medical record data.

Images